Theoretical LED limit.

[coolwaters]

by using LEDs can it reach the Theoretical luminous efficacy of 683.002lm/watt?

 

[jtr1962]

Only with monochromatic green at 555 nm. For white light made by mixing red, green, and blue emitters the maximum is roughly 400 lm/W at a CCT of about 4000K (it's less at higher or lower CCTs).

 

[paulr]

Think we'll get there? Or even to 200 l/w?

 

[coolwaters]

200l/w would be great...

but 400 is too unreal right now...just imagine the power saved.

 

[2xTrinity]

Think we'll get there? Or even to 200 l/w?

If they could make the green junctions as efficient as the best blue ones presently are, we would already be at 200 lm/W for RGB (the best blue and red LEDs I believe are pushing 50% radiant efficiency). I believe the poor performance of the green is one of the main reason why phosphor LEDs are the main way in which white light is generated. IMO a multi-emitter light with red, amber, green, and blue should produce much better results, and with the potential to arbitrarily vary the CCT at will.

 

[jtr1962]

Think we'll get there? Or even to 200 l/w?

My own personal feeling is that we will get to about 75% or 80% efficiency within a decade. As 2xTrinity said, the main problem right now is that green emitters are lagging red and blue ones, both of which are approaching 50% efficiency. I feel eventually this problem will be solved. There just hasn't been much incentive to improve green emitters as exists with blues and reds. Blues are used to make whites, so obviously the white LED efficiency race has pushed their development. Reds are used in traffic lights and automobiles. Here again there has been a huge incentive to make them more efficient. The main use for green right now is traffic lights. The present ones may be less efficient than the reds, but they are "good enough" for this application, using only 10% of the power of the incandescents they replace. There is less incentive to improve them because the incremental power savings are small. Reds needed to be improved faster since the eye is less sensitive to red. To make reds of the same brightness and power consumption as greens, they needed to have about three times the conversion efficiency. However, once blue plus phosphor whites start getting near their theoretical limits of roughly 200 to 225 lm/W, the only path forward will be with RGB whites. This will give the needed incentive to improve the greens.


Anyway, I do think 75% to 80% (300 to 320 lm/W) is likely to happen. I even feel 90% efficiency (360 lm/W) is a remote possibility. These figures are four to five times the efficiency of present-day CFLs, and about 20 times better than incandescents. Amazing to think that in time a 5 watt LED may well do what a 100 watt incandescent lamp does.

 

[paulr]

:wow: Thanks! I'd been wondering as well what the limits are for phosphor whites. Looks like we're already halfway there. :thumbsup:

 

[SemiMan]

I just love how misinformation gets spread in public forums. It is truly a thing of wonder......

The best shipping blue LEDS are about 35% efficient, maybe 40 and that is really pushing it. Remember the Blue Cree announcement is a maximum of 42 lumens and the wavelength is not specified, but there blue bin goes up to 485nm where 42 lumens would be about 35% efficient.

Reds are currently no where near 40% efficient. They are more in the 20% range.

Why not RGB whites:

- 3 monochromatic sources makes for a terrible light source with poor CRI. It is not the ideal light source by any stretch where color quality is concerned. A blue excited phosphor LED can have much better CRI.

- RED led output currently drops very quick with temperature

- It is not at all easy to control the color output of an RGB light without feedback and control mechanism versus blue phosphor leds where it is somewhat fixed after it is made.

- It is possible to make reasonably efficient green emitters. However, for the reasons above, there is no driving force yet for high volumes of green emitters as RGB white is not yet a market driver. Remember things like thin film and the other advances that apply to blue leds apply to green leds as well.

- One thing that is missing is the ability to create green efficiently in the 550nm range. Most Greens are in the 530 range.

Semiman

 

[jtr1962]

I just love how misinformation gets spread in public forums. It is truly a thing of wonder......

The best shipping blue LEDS are about 35% efficient, maybe 40 and that is really pushing it. Remember the Blue Cree announcement is a maximum of 42 lumens and the wavelength is not specified, but there blue bin goes up to 485nm where 42 lumens would be about 35% efficient.

Reds are currently no where near 40% efficient. They are more in the 20% range.

We were talking about the best lab efficiency, not production efficiency. BTW, Cree does have a production minimum 30 mW bin for their EZR blue chip. At a typical 64 mW input power this translates to a minimum efficiency of 47%. Granted, production power blue LEDs are still lagging behind this a bit, but in the lab they've already reached about 50%. Cree's 129 lm/W at 350 mA lab result represents an overall conversion efficiency of roughly 39%. Since blue to white phosphor conversion is at best 80% efficient at the wavelengths in question, this implies a blue chip with an efficiency of at least 49% at 350 mA.

As for reds, I'm aware of some which get roughly 42 lm/W at 350 mA. This was several years ago so we may be doing better now. Raw conversion efficiency obviously depends upon the dominent wavelength. If we assume 625 nm then we're around 19%, close to what you said. I'm pretty sure we have 5mm reds which are doing about twice as well. As with greens, there hasn't been much incentive the last few years to improve reds as the existing ones are plenty good enough for their intended application.

I think it'll be a few more years before we can economically solve the problems with RGB whites which you mentioned. By then phosphor whites will probably be pushing their limits, so there will be added incentive to improve greens and reds.

 

[SemiMan]

I forgot about the new Blue EZR family. It is exceptionally efficient. Thank you for pointing that out. Unfortunately, reds have not gotten much better over the last several years...maybe a 20-25% increase but not the 2-3x we have seen with blue. AlInGaP is a much more mature technology than InGaN. I had seem some promises over the years on less drop off with temperature but have yet to see anything hit production that seems to offer much of a benefit.

Semiman

 

[laserblue]

What we are aiming for is electrical efficiency or wall plug efficieny not luminous efficacy. we should build leds that are electrically super efficient.

 

[SteveDavis]

The numbers that I've seen for a pure white light source put maximum lumens/watt at around 240, but obviously, this varies depending on the source output spectrum. RGBA white can produce decent CRI, and will probably be more efficient than phosphors eventually, however, as was mentioned earlier, control is very difficult without an optical sensor. This is because color and intensity binning for multi color LED's is extremely difficult, and because AlInGaP and InGaN dice behave differently at higher temperatures, and age differently.

From what I've seen out of the LED vendors, I would expect white phosphor based LEDs to asymptotically approach about 200lm/W. The maximum wattage of a single emitter will continue to go up for the foreseeable future, and the number of emitters per package will also go up.

Those are just my personal opinions.

 

[chris_m]

From what I've seen out of the LED vendors, I would expect white phosphor based LEDs to asymptotically approach about 200lm/W. The maximum wattage of a single emitter will continue to go up for the foreseeable future

Emitter wattage increasing kind of follows from efficiency increasing. If maximum theoretical is 240lm/W, then going from 100lm/W (where we are now) to 170lm/W (2-3 years away? or will current progress slow down before getting near there?) halves the heat output, so theoretically allows you to run twice the input power for a given amount of heating. In fact we're not that far off the point where improvements in efficiency will be more important because of the effect they have on heat output.

 

[JohnR66]

One thing I see looking at other light sources, there have been minor improvements in efficiency over the years but then they hit a brick wall and can do no better. I doubt there will be any more big jumps.

What I see:
Small improvement to efficiency (better improvements to warm white colors)
Improved color consistancy (less color binning)
Higher power single LEDs
Lower cost.

LEDs will be really mainstream then.

 

[SemiMan]

Emitter wattage increasing kind of follows from efficiency increasing. If maximum theoretical is 240lm/W, then going from 100lm/W (where we are now) to 170lm/W (2-3 years away? or will current progress slow down before getting near there?) halves the heat output, so theoretically allows you to run twice the input power for a given amount of heating. In fact we're not that far off the point where improvements in efficiency will be more important because of the effect they have on heat output.

There are some other factors to look at. Currently, when you increase the current in the LED, the efficiency drops. We either need technology to fix that or go with bigger dies. OSRAMS latest die is 2mm squared versus the 1mm that has traditionally been used. That will result in a cost increase. How much we do not know yet.

Semiman

 

[Canuke]

- 3 monochromatic sources makes for a terrible light source with poor CRI. It is not the ideal light source by any stretch where color quality is concerned. A blue excited phosphor LED can have much better CRI.

That is due to the wide emission band of the phosphor, relative to the narrow blue spike of the source emitter.

Would CRI increase with wider-band emitters? For example, I have seen some blue-green LED's which have a really fat spectrum, emitting a wash of light from 520 to 480nm. It's color to the eye is visibly less saturated than narrower-band emitters. I think Craig has a few spectra from such wideband LED's.

An RGB light source using such emitters would lose the ability to do very saturated colors, but could still be tuned to a wide range of colors useful for general lighting and mood. The resulting spectrum would be a lot more smooth and even, being the sum of three wide RGB hills instead of three narrow RGB spikes.

 

[Curious_character]

Thanks for the info on a most interesting thread.

What I'd like to know is what kind of efficiency current and future LEDs get in terms of the fraction of input power that's radiated as light at any wavelength. The remaining power is what we have to take care of by heat sinking.

For example, let's suppose that we can keep the LED temperature within specs while getting rid of 3 watts of heat. I think most flashlights today can do that. Using the definition of efficiency I described, if an LED is

25 percent efficient, then we can drive it with an input power of 4 watts.
50 percent efficient, then we can drive it with an input power of 6 watts.
75 percent efficient, then we can drive it with an input power of 12 watts.
90 percent efficient, then we can drive it with an input power of 30 watts.

Without exceeding the maximum LED temperature.

So it looks to me like the total lumens you can get out of an LED is determined not only by the luminous efficiency (lumens/watt), but also by your heat sinking capability (including the thermal resistance of the LED case) and the overall efficiency (watts of light of any wavelength out per watt in). As you can see, the overall efficiency in terms of total watts of light of any wavelength per watt of input power is an important factor in determining how much input power we can apply and, in conjunction with the luminous efficiency, how many total lumens we can get out of a particular LED. Even with a moderate luminous efficiency, 30 watts of power will produce a lot of light. Of course, if we want to drive a highly efficient LED at 30 watts in a flashlight, the problem then becomes how big a battery we can fit into the light and/or how short a run time we'll tolerate.

c_c

 

[SteveDavis]

Canuke: Can you post a link to a datasheet on one of those wide spectrum non-phosphor products? I have never heard of such a thing.

 

[JohnRobHolmes]

Thanks for the info on a most interesting thread.

What I'd like to know is what kind of efficiency current and future LEDs get in terms of the fraction of input power that's radiated as light at any wavelength. The remaining power is what we have to take care of by heat sinking.

For example, let's suppose that we can keep the LED temperature within specs while getting rid of 3 watts of heat. I think most flashlights today can do that. Using the definition of efficiency I described, if an LED is

25 percent efficient, then we can drive it with an input power of 4 watts.
50 percent efficient, then we can drive it with an input power of 6 watts.
75 percent efficient, then we can drive it with an input power of 12 watts.
90 percent efficient, then we can drive it with an input power of 30 watts.

Without exceeding the maximum LED temperature.

So it looks to me like the total lumens you can get out of an LED is determined not only by the luminous efficiency (lumens/watt), but also by your heat sinking capability (including the thermal resistance of the LED case) and the overall efficiency (watts of light of any wavelength out per watt in). As you can see, the overall efficiency in terms of total watts of light of any wavelength per watt of input power is an important factor in determining how much input power we can apply and, in conjunction with the luminous efficiency, how many total lumens we can get out of a particular LED. Even with a moderate luminous efficiency, 30 watts of power will produce a lot of light. Of course, if we want to drive a highly efficient LED at 30 watts in a flashlight, the problem then becomes how big a battery we can fit into the light and/or how short a run time we'll tolerate.

c_c

Excellent post.

 

[Canuke]

Canuke: Can you post a link to a datasheet on one of those wide spectrum non-phosphor products? I have never heard of such a thing.

Craig has two graphs of such "broadband" LED's; go to this page http://ledmuseum.home.att.net/specx21.htm

...and search for "broadband" on that page. One of them is a Nichia NLPB320AS, and there is another Panasonic LED mentioned also. From looking at those spectra, I get the impression "broadband" emitters happen only around the blue-green area, so perhaps that's a "happy accident" of that emission band/chemistry.

If you fish around on Craig's site, you'll see some spectra from "non-phosphor whites" and a few emitters classified as "blue" also show fat bands.

Most of the emitters he labelled as "broadband" also seem to be pretty old and/or low power/indicator light types.

I have two items, one a color-changing display stand, and another is a set of LED ice cubes, which have green emitters that are not very saturated to my eye, and seem to have a wide emission band as seen with a CD (I don't have a spectrum analyzer).